aryyanthakrr/Kepler-Reasoning-7B

TEXT GENERATIONConcurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 3, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Kepler Reasoning 7B is a 7.6 billion parameter model developed by aryyanthakrr, created by merging Qwen2.5-Coder-7B-Instruct and Qwen2.5-Math-7B-Instruct using the SLERP method. This model is specifically optimized for strong local coding and mathematical reasoning tasks. It features a 32768 token context length, making it suitable for complex problem-solving in these domains.

Loading preview...

Kepler Reasoning 7B: Merged for Code and Math

Kepler Reasoning 7B is a 7.6 billion parameter language model developed by aryyanthakrr, specifically engineered for enhanced local performance in coding and mathematical reasoning. It was created by merging two specialized Qwen models: Qwen2.5-Coder-7B-Instruct and Qwen2.5-Math-7B-Instruct, utilizing the SLERP (Spherical Linear Interpolation) merge method.

Key Capabilities

  • Coding Assistance: Provides support for various coding tasks.
  • Python Problem Solving: Excels at solving problems specifically within the Python programming language.
  • Mathematical Reasoning: Capable of handling complex mathematical problems.
  • Algebra and Word Problems: Designed to address both algebraic equations and text-based math challenges.
  • Local Inference: Optimized for efficient local deployment, including GGUF quantization.

Intended Use Cases

This model is ideal for developers and researchers requiring a robust local solution for:

  • Generating and debugging code snippets.
  • Solving programming challenges, particularly in Python.
  • Assisting with mathematical computations and logical deductions.
  • Tackling algebra and word problems in educational or technical contexts.

Limitations

As a 7B parameter model, Kepler Reasoning 7B is designed for local inference and specialized tasks. While highly capable in its target domains, it is not expected to outperform larger, frontier cloud-based models in general-purpose AI benchmarks.